Table 3 Recall and precision increase for candidate search filter pairings
From: Development and validation of paired MEDLINE and Embase search filters for cost-utility studies
Filter pair | Target CUAs identified | Total yield (2010 only) | ||||
|---|---|---|---|---|---|---|
N | Recall (95% CI) | Number of hits | Precision increase (compared to existing strategy) | |||
MEDLINE | Embase | Deduplicated | ||||
Existing NICE strategy | 370 | 100.0% (99.0 to 100.0%) | 56,398 | 108,133 | 111,467 | – |
Sensitive | 370 | 100.0% (99.0 to 100.0%) | 13,262 | 20,277 | 22,085 | 5.0x |
Precise | 361 | 97.6% (95.4 to 98.9%) | 3,337 | 3,279 | 4,966 | 21.9x |
Free-text only | 353 | 95.4% (92.8 to 97.3%) | 1,675 | 3,137 | 3,208 | 33.2x |